ANN Method for Control of Robots to Avoid Obstacles

Abstract

The avoidance of obstacles placed in the workspace of the robot is aproblem which makes controlling them more difficult. The known avoidance methodsused for the robots control are based on bypass trajectory programming or on usingthe sensors that detect the position of the obstacle. This paper describes a method oftraining industrial robots in order for them to avoid certain obstacles in the workspace.The method is based on the modelling of the robot’s kinematics by means of anartificial neural network and by including the neural model in the robot’s controller.The neural model simulates the robot’s inverse kinematics, and provides the jointcoordinates, as referential values for the controller. The novelty of the method consistsin the deliberately erroneous training of the network, so that, when programming adirect trajectory in the workspace, the robot avoids a known obstacle.